2011-03-20T02:00:30 what's a zombie process? 2011-03-20T02:02:48 http://www.brighthub.com/computing/linux/articles/79186.aspx 2011-03-20T02:02:59 seems to be a pretty good description 2011-03-20T02:04:49 okay cool 2011-03-20T02:05:04 my process thing is creating them 2011-03-20T02:05:31 they die when the parent program exits, but i want to be able to kill them off before that 2011-03-20T02:06:01 *** ravic has quit IRC (Remote host closed the connection) 2011-03-20T02:08:18 http://www.makelinux.net/alp/026.htm might help you with that 2011-03-20T02:08:56 assuming whatever your process thing is, is in c/c++ 2011-03-20T02:12:27 it sort of is 2011-03-20T02:12:33 i need to do the whole wait thing 2011-03-20T02:13:04 but their example just declares an int then says wait(that int) but doesn't do anything with the int before? wtf? 2011-03-20T02:13:28 the 3.4.2 section 2011-03-20T02:13:42 is there a "alt tab" for switching desktops? 2011-03-20T02:13:55 no, it says wait(address of that in) 2011-03-20T02:14:01 it's an output value 2011-03-20T02:14:25 the int is filled with the child status 2011-03-20T02:14:34 see http://linux.die.net/man/2/wait 2011-03-20T02:26:44 hmmm i don't know how to get the process ID :P 2011-03-20T02:26:55 each process has FILE in,out; 2011-03-20T02:27:31 pid of processes that you are creating, or pid of processes you are waiting on? 2011-03-20T02:27:58 I have a process struct 2011-03-20T02:28:24 i didn't write the start function, it's that same one i've been fixing up for a while 2011-03-20T02:28:25 how did you create the process? 2011-03-20T02:29:25 either way, you don't need the pid to wait 2011-03-20T02:29:37 what do i need? 2011-03-20T02:29:59 i don't get how just the address of some random int tells you the status of some other running process :\ 2011-03-20T02:30:18 surely that int needs to be linked to the process in some way? 2011-03-20T02:30:33 the return value of wait is the pid 2011-03-20T02:30:44 and it won't be running anymore by that point 2011-03-20T02:31:14 err what, but i might have a bunch of these processes running 2011-03-20T02:31:21 how does it know which one i'm calling wait on? 2011-03-20T02:31:57 wait() waits until any child process finishes up 2011-03-20T02:32:21 if you want to wait for a specific process then there is a version of wait which takes more arguments 2011-03-20T02:32:46 let me see where i get with just wait 2011-03-20T02:32:48 cheers 2011-03-20T02:32:53 janzert: thanks too 2011-03-20T02:33:07 wait() basically says "I can't continue until my children have finished processing", so tell me when one finishes 2011-03-20T02:34:28 if you just want to wait until all children have finished then put wait in a while loop 2011-03-20T02:36:29 the children still seem to become zombies until my parent process exits 2011-03-20T02:40:19 what are you doing? 2011-03-20T02:42:55 I mean, the absolute bare-bones thing to do is while(wait() != -1); but that doesn't do any error checking or anything 2011-03-20T02:44:44 no error checking is fine 2011-03-20T02:45:47 that doesn't compile :P 2011-03-20T02:46:30 oh, well wait needs to take that int *status argument 2011-03-20T02:47:23 still wont compile :P 2011-03-20T02:47:38 http://pastebin.com/XDApjjwB 2011-03-20T02:47:52 that's my function, the for loop is where i'm trying to get a process to end properly 2011-03-20T02:50:35 i've tried the wait in any (un)obvious place 2011-03-20T02:51:03 what's the error? 2011-03-20T02:52:19 well, you don't want to wait until after you have finished talking to the proccess 2011-03-20T02:53:17 remember, the wait is for when you are waiting for the child to finish, if the child is trying to write to you then waiting will just result in a deadlock 2011-03-20T02:55:20 it doesn't result in a deadlock 2011-03-20T02:55:24 it runs fine :\ 2011-03-20T02:55:42 i don't think wait is waiting for anything 2011-03-20T02:55:50 http://pastebin.com/NCDazQr8 2011-03-20T02:56:04 that's the error i get for "while(wait(*test) != -1);" 2011-03-20T02:57:21 my bad, http://pastebin.com/PufT2NJb for &test 2011-03-20T02:57:28 i had tried both, both hated me 2011-03-20T02:59:19 did you #include 2011-03-20T03:00:24 either way, this seems like something the process class should be taking care of 2011-03-20T03:00:45 *shrug* 2011-03-20T03:02:48 it would be, i want to work out how to first :P 2011-03-20T03:02:52 i'll play some more 2011-03-20T03:04:27 also, why do you have a vector of processes when you only have one child process at a time? 2011-03-20T03:05:03 well, i currently don't have a way to cleanly close and reopen the same process object 2011-03-20T03:05:07 but i'm working on that now 2011-03-20T03:05:19 and i seem to have got somewhere by including sys/wait.h :D, cheers 2011-03-20T03:05:28 i'll keep playing now till it's clean 2011-03-20T03:05:56 is sys/wait.h on windows too? 2011-03-20T03:06:42 not that it matters too much to me now, but i might as well try and keep it working on everything now 2011-03-20T03:08:04 no, unless you something that enumlates a unix environment like cygwin or minggw 2011-03-20T03:08:31 I don't know anything about windows programming though 2011-03-20T03:08:54 i use mingw usually 2011-03-20T03:09:09 mingw is nice 2011-03-20T03:22:29 okay i can now reuse it :) 2011-03-20T03:22:31 cheers again 2011-03-20T03:26:36 *** Stocha has joined #aichallenge 2011-03-20T03:39:36 The problem with using both friendy and ennemy ant, is that it is even slower :/ 2011-03-20T03:40:23 If you go with counting nearby ants, give a quick efficient algorithm with every starter kit :p 2011-03-20T03:40:37 though that really is intuitive 2011-03-20T03:41:30 but that would allow blocking unless it is either statistical. Or it always allow to reach contact point some way or others 2011-03-20T03:41:49 at distance 1, you can maximum be 3 vs 3 i think 2011-03-20T03:42:19 so worst blocking scenario would be a 1 corridor like in my gif 2011-03-20T03:42:27 .b. 2011-03-20T03:42:35 bbb 2011-03-20T03:42:39 .a. 2011-03-20T03:42:47 ah no 2011-03-20T03:42:50 %b% 2011-03-20T03:42:52 bbb 2011-03-20T03:42:56 %a% 2011-03-20T03:42:59 %.% 2011-03-20T03:43:01 %.% 2011-03-20T03:43:14 so it is 2 vs 3 here 2011-03-20T03:43:36 if two vs three always die that a is blocked there. 2011-03-20T03:44:41 Life point or statistical will solve that. 2011-03-20T03:45:14 But then we have double slow counting ennemy and friendly units. And have complex state (life point) or tricky testing (statistical) 2011-03-20T03:46:08 I understand that everyone wants complex fighting. I really think we have to something straight forward and fast to calculate for it to be good. 2011-03-20T03:48:32 Stocha: how is it slower? 2011-03-20T03:58:48 iterators are really just structs inside structs yeah? 2011-03-20T03:59:17 why not template the iterator with the type it is point to, then you can use a template to create an iterator for your template type 2011-03-20T04:03:07 it, taking a template Game, 2011-03-20T04:03:07 Game::NStrategy doesn't work, but 2011-03-20T04:03:08 NStrategy would 2011-03-20T04:06:38 oh, but then you'd need to name them all differently 2011-03-20T04:06:40 that doesn't help me 2011-03-20T04:09:47 *** amstan has quit IRC (Ping timeout: 250 seconds) 2011-03-20T04:12:23 fast is : logic bit operation and simple addition. Anything else is slow. Then there is the consideration of Complexity against the total number of ants in the worst case. 2011-03-20T04:12:52 fast is O(n) (with low multiplicator) where n is total number of ant present on the map. 2011-03-20T04:13:12 Ant dies in sames square falls into that categorie. 2011-03-20T04:13:27 Stocha: it doesn't really matter provided it's fair 2011-03-20T04:13:32 the simple rules are boring 2011-03-20T04:13:43 you're one of the few people i know of who think they aren't 2011-03-20T04:16:11 i don't think they aren't. I just think it might be too soon 2011-03-20T04:16:23 and that they have to be lighting fast to compute and testcase. 2011-03-20T04:16:38 then i will be happy to have whatever complexity. 2011-03-20T04:16:42 *** delt0r___ has joined #aichallenge 2011-03-20T04:16:43 and big tree solvers and such. 2011-03-20T04:17:33 but if it's hard to compute, to undestand the spec, and to implement an unoptimized version, then i'm really worrying. 2011-03-20T04:17:51 because that will not be fun for most casual people willing to participate. 2011-03-20T04:18:21 *** delt0r_ has quit IRC (Ping timeout: 255 seconds) 2011-03-20T04:18:43 People will want to solve interesting games. Not to debug / optimize their code for month, before starting to really work on the ai challenge itself. 2011-03-20T04:19:32 So if you think complexity is good, make it linear in the number of ant, and easy to understand/ check for rightness if implementation. 2011-03-20T04:22:43 it is linear in the number of ants, because you only need to check a constant number of map positions per ant 2011-03-20T04:27:07 constant time would be even better :p 2011-03-20T04:28:32 well, that would be doing no attacks what so every 2011-03-20T04:28:43 *ever 2011-03-20T04:29:03 In my case it was not linear : i would find the ennemy ant by checking the shape around. Then iterate over the enemies to count them :p 2011-03-20T04:29:30 I made a post about a rock/scisor/paper combat for ants. 2011-03-20T04:29:42 but the post got lost when it asked me to login again. 2011-03-20T04:30:09 I think the success of tron and planet wars as AI challenge, came from the simplicity of thoses games. 2011-03-20T04:30:20 You had straight forward rules, very easy to understand. 2011-03-20T04:30:22 all the methods are constant time per ant, if that's what you mean 2011-03-20T04:30:48 Then you could put all your mind into solving them. Not debugging for years, or trying to understand what the heck those specs means. 2011-03-20T04:31:08 no constant time doesn't depends on ant numbers. 2011-03-20T04:31:12 people manage to make sort O(n**2) instead of O(n log n) all the time too, it's not the fault of sorting ;) 2011-03-20T04:31:37 You can sort in linear time also. 2011-03-20T04:31:42 Stocha: the current methods are constant time per ant 2011-03-20T04:31:47 Think how child sort out cards on the table. 2011-03-20T04:32:06 with a vision radius of 6, a constant number (21) of map positions need to be checked 2011-03-20T04:32:22 You take cards one by one, you put them "at their right place on the table", 1 place for each 52 cards, then you stack them one by one, sorted. 2011-03-20T04:32:36 yes, see radix sort 2011-03-20T04:32:45 or bucket sorting 2011-03-20T04:33:14 of course they come with other limitations though 2011-03-20T04:33:32 As i use bit maps, i have time linear in the number of long it take to bitmap the map, for solving collisions. 2011-03-20T04:33:43 doesn't work on things which you can't compare with the bit representation 2011-03-20T04:33:50 yeah that O(***) is meaningless i guess :p 2011-03-20T04:33:54 (re limitations) 2011-03-20T04:34:10 what are the sort of thing you can't compare with bit representations ? 2011-03-20T04:34:14 i don't get it :p 2011-03-20T04:34:33 How do you compare them then ? With quantical representation ? :p 2011-03-20T04:34:48 say a simple implementation of rational numbers 2011-03-20T04:34:57 what ?? 2011-03-20T04:35:40 You mean complex bit comparison then :p 2011-03-20T04:35:40 if you store a rational number as a pair (numerator, denominator) sorting their bit representation wont sort by value 2011-03-20T04:35:55 then radix sorts wont work 2011-03-20T04:36:03 Stocha: see http://en.wikipedia.org/wiki/Radix_sort if you're not familiar with radix sort 2011-03-20T04:36:15 what if you keep them by pair of reduced numbers ? 2011-03-20T04:36:16 you are into general comparision sorts which have a optimal bound of O(n log(n)) 2011-03-20T04:36:19 I'm pretty sure it explains the limitations sigh is refering to 2011-03-20T04:36:32 Stocha: how will reduced numbers help? 2011-03-20T04:37:17 for example, the following is in sorted order in reduced repr: (1,3), (1,2), (2,3), (1,1), (2,1) 2011-03-20T04:37:23 you have only one representation for any number then ? 2011-03-20T04:37:24 anyway have fun, 4:30am so I really should go to bed ;) 2011-03-20T04:37:46 Stocha: yes, but that representation doesn't work for radix sort 2011-03-20T04:38:13 you can make a representation for rationals that work for radix sort, but that isn't good for adding and multiplying them 2011-03-20T04:38:15 let's be discussing about the game then, as it appears i'm somewhat lacking in algorithmic culture :p 2011-03-20T04:38:47 I hate numbers since i heard of uncountable sets anyway. 2011-03-20T04:39:16 well I was only talking about rationals, which are countable 2011-03-20T04:39:19 especially when you try to trick me into "infinit set of whatever" 2011-03-20T04:39:56 that's still bigger than a 64 bits operand. 2011-03-20T04:40:23 one number being an infinite string with ununderstable rules for comparison :p 2011-03-20T04:40:52 so let's be on about the need for simplicity. 2011-03-20T04:41:10 Anywhere i look i see people asking for "complex interesting ant game" :'( 2011-03-20T04:41:16 what can i do about that ? 2011-03-20T04:41:19 why do you want basic? 2011-03-20T04:41:37 Stocha: by complex they mean they want the emergent strategies to be complex, not the rules 2011-03-20T04:41:44 Why do i beleave it has to be easy to implement and test for ? like less than 4 days to have an engine work fast. 2011-03-20T04:41:57 without a good battle resolution i am worried the game would just be fastest to find food and not be stupid 2011-03-20T04:42:12 you don't need to make your own engine 2011-03-20T04:42:28 You have to, if you want to build a solver 2011-03-20T04:42:37 lol, no you don't 2011-03-20T04:42:47 you won't be able to make a tree of anything, if you don't have rules for calculating the next nodes. 2011-03-20T04:42:50 you need to emulate some things, but you don't need your own working engine 2011-03-20T04:42:57 yes, but that isn't an engine 2011-03-20T04:42:57 same goes with montecarlo trees. 2011-03-20T04:43:12 Yeah, thoses things need to be easy to emulate then. 2011-03-20T04:43:20 they are 2011-03-20T04:43:24 less than a full day work for unoptimized. 2011-03-20T04:43:28 you don't just go with the most trivial option 2011-03-20T04:43:31 you go with the most interesting 2011-03-20T04:43:35 and with few risk of getting wrong. 2011-03-20T04:43:43 that's stupid 2011-03-20T04:43:57 planet wars is such a game. 2011-03-20T04:43:59 so is tron. 2011-03-20T04:44:05 (they are even simpler i guess but well) 2011-03-20T04:44:06 planet wars wasn't so simple 2011-03-20T04:44:14 and there were lots of little fiddly bits here and there 2011-03-20T04:44:15 what rues are hard to test ? 2011-03-20T04:44:23 rules 2011-03-20T04:44:30 there was a huge fuss over battle resolution for that 2011-03-20T04:44:35 planet wars has annoying battle resolution, and a few bugs sprang up during the comp 2011-03-20T04:44:44 and the tiniest change in things caused a huge change in optimal strategies 2011-03-20T04:44:46 You have planet regeneration, and then ships, and then ships order 2011-03-20T04:44:55 it was very hard to make any visible progress after a while as well 2011-03-20T04:45:09 progress at what ? bot strength ? 2011-03-20T04:45:14 yes 2011-03-20T04:45:20 why is it that a bot totally overclassed the others then ? 2011-03-20T04:45:25 most people seemed to hit a brick wall after a while 2011-03-20T04:45:31 because mega1 is mega 2011-03-20T04:45:36 that would go for any challenge. 2011-03-20T04:45:40 he pantsed everybody 2011-03-20T04:45:51 not like that, tron had no such wall 2011-03-20T04:46:01 and i expect people to find a lot more to write stuff for with this than pw 2011-03-20T04:46:01 He solved it better than the other. He won the challenge. 2011-03-20T04:46:18 ah, tron didn't have the wall, okay. you have a point then. 2011-03-20T04:46:28 didn't people had fun with planet war ? 2011-03-20T04:46:30 that's another reason for now just having the most basic game mechanics with a straight forward strategy for micro and macro, it gives people obvious things to improve, and they can see clear results 2011-03-20T04:46:45 they went and said : let me regret to have applyed ? :p 2011-03-20T04:46:47 planet wars was an awesome little problem 2011-03-20T04:46:56 but i personally don't think it made for a great ai game 2011-03-20T04:47:05 it was awesome to watch as well etc. 2011-03-20T04:47:13 i just feel it was lacking in the way of decent strategies 2011-03-20T04:47:27 Still it fast to compute state for. 2011-03-20T04:47:30 that's good. 2011-03-20T04:47:40 and you can build a working state transition in a day. 2011-03-20T04:47:45 that's good too. 2011-03-20T04:47:49 you can do that with the current options 2011-03-20T04:47:54 i don't see that as necessary 2011-03-20T04:48:08 and you can read the spec on time, and go : mm i think i understand what the game is, and how to implement it :p 2011-03-20T04:48:11 if you're not willing to spend a bit of time making your bot good, you shouldn't expect to win 2011-03-20T04:48:18 i want to play around with cooperation in n-player games 2011-03-20T04:48:30 I want to make the bot, not debug the spec of the game :p 2011-03-20T04:48:38 i want them to be crystal clear. 2011-03-20T04:48:48 they are clear 2011-03-20T04:48:56 that they are not at this point. 2011-03-20T04:49:08 it's still in development 2011-03-20T04:49:14 the actual rules are clear, the specs may not be 2011-03-20T04:49:47 but for the fact that ant receive a dir order and cant walk on water. and one food spawn one ant if only one color is nearby and disapear if more than one is nearby 2011-03-20T04:49:52 that is what crystal clear is :) 2011-03-20T04:50:29 <_flag> There's small subtle rules, but the basic rules are fairly clear 2011-03-20T04:50:29 and that if more than one ant occupy a square, all ant in that square disapears; That's clear too. 2011-03-20T04:51:03 there's small unatural hard to compute/ test/ implement ... hard to understand rules. 2011-03-20T04:51:39 the idea is we want attacks 2011-03-20T04:51:43 but all them are unfinalized also. 2011-03-20T04:51:51 and your "favourite rule" isn't attacks, it's removing collided ants 2011-03-20T04:51:52 well, the only thing in regard to the bots that matters here is battle resolution 2011-03-20T04:52:05 which is why it is being discussed now 2011-03-20T04:52:05 you other proposed method is random 2011-03-20T04:52:10 which is laughably less clear 2011-03-20T04:52:17 The only thing that matters so far, is exploration, and food gathering 2011-03-20T04:52:27 no, attacks matter 2011-03-20T04:52:31 that is how you get points 2011-03-20T04:52:41 <_flag> You have a set of ants, you move them around to find food for more ants, and your ants die if they have more or equal enemies around them than friends. There was some odd aspects of planet wars battle resolution dealing with neutral planets, but that didn't stop anyone from saying "I'll just pick the planet with the smallest number of ships and send 50% of my ships there". 2011-03-20T04:52:41 i doubt that your fighting rules make it that the one that get the exploration and food gathering better won't win. 2011-03-20T04:53:25 that will be an important component, but not the only one 2011-03-20T04:53:29 nor do i want it to be 2011-03-20T04:53:31 <_flag> Similarly, if you want to be the best, you have to understand all the intricacies of the game. Otherwise, there's nothing stoping you from making a bot that moves around randomly 2011-03-20T04:54:03 that's what i did so far :p 2011-03-20T04:54:04 there will be a range of levels of bots for people to compete against 2011-03-20T04:54:08 a bot that moves randomly. 2011-03-20T04:54:24 but why make the game so simple that anyone willing to put a tiny bit of effort is going to be fairly noisely fighting it out near the top? 2011-03-20T04:54:35 Well i guess, that if i'm right it won't be a real problem after all. 2011-03-20T04:54:44 ? 2011-03-20T04:54:45 top players will be solving the food gathering thing. 2011-03-20T04:54:58 and stockpiling at borders. 2011-03-20T04:55:00 i've already got my food gathering algorithm worked out 2011-03-20T04:55:07 that's how every rules end anyway :p 2011-03-20T04:55:12 that's how boring i think it is 2011-03-20T04:55:14 <_flag> Food gatherthing isn't a problem that takes months to solve on its own 2011-03-20T04:55:35 i have trouble believing it could be optimal in such a few effort :p 2011-03-20T04:55:52 The game is about that, ant life is about that :) 2011-03-20T04:56:12 find the border of each food square, then paint outwards and find "closest ants" repeatedly moving ties until a player will win 2011-03-20T04:56:26 They have been studied for years, for that. How they find food and bring it back. 2011-03-20T04:56:31 the only difficult is possibly switching up the order of moving ants to food in order to make sure you gaurantee getting more 2011-03-20T04:56:41 but that's a little bit of play work, not a whole contests worth 2011-03-20T04:56:43 They have a nice solving of that game too. ants. 2011-03-20T04:57:11 So you have the go to the border problem then. 2011-03-20T04:57:18 that's easy 2011-03-20T04:57:22 even with simple dieonsamesquare you have that. 2011-03-20T04:57:23 i paint to find that too 2011-03-20T04:57:33 or i can store the translate amounts for a border 2011-03-20T04:57:40 you always says you solved it. but i have no evidences of that. 2011-03-20T04:57:57 it's not completely worked out, but pretty much 2011-03-20T04:58:10 If you solved it in a day then you 'll win or the game is too simple. 2011-03-20T04:58:16 i need to go through and implement and iron through a few things, but it's certainly not a contest making problem 2011-03-20T04:58:37 you maybe right 2011-03-20T04:58:51 i have to admit that. 2011-03-20T04:59:17 though i doubt it at this point, it may be that you will be right in the end. but what is the probability of your being right about it i wonder. 2011-03-20T04:59:40 if you have the enemies>=friends then the best strategy isn't necessarily to fight something as fast as possible, you will want to dance around each other and try to improve your death ratio 2011-03-20T04:59:46 can you imagine how cool that will look? 2011-03-20T05:00:43 The rule i proposed wall all about enemies>friends ratio. 2011-03-20T05:00:52 although it was statisticall 2011-03-20T05:00:59 it could also be life based. 2011-03-20T05:01:31 Stocha: are you aware that we've pretty much got a new favourite battle option? 2011-03-20T05:01:32 I still stated that i had this feeling that it was too soon. 2011-03-20T05:01:46 the one where your ant dies if it has at least as many ants in its battle radius than friends? 2011-03-20T05:01:53 rather than the power option 2011-03-20T05:02:03 Let's first go with a challenge about solving the simple ant wars game. Then next challenge you had interesting fighting based on the feed back of 5000 users ? 2011-03-20T05:02:06 enemies > friends is implemented as "--attack support" in the engine 2011-03-20T05:02:12 sigh: with McLeopold's aggreance, i think you could almost make that the new default 2011-03-20T05:02:33 meh, I'll make it the new default now 2011-03-20T05:02:40 easy 2011-03-20T05:02:40 <_flag> sigh: enemies > friends or enemies >= friends? 2011-03-20T05:02:48 current > 2011-03-20T05:02:54 what is your opinion, _flag 2011-03-20T05:03:02 sigh: 1vs1 doesn't die then? 2011-03-20T05:03:12 1vs1 does die, 1 vs has no friends 2011-03-20T05:03:16 1v1 2011-03-20T05:03:19 <_flag> Oh 2011-03-20T05:03:22 <_flag> Okay nevermind then 2011-03-20T05:03:27 okay sure 2011-03-20T05:03:30 leave it at that 2011-03-20T05:03:40 maybe even make a comment note of that 2011-03-20T05:03:41 is the ant is own friend or not ? 2011-03-20T05:03:48 that should be stated once and for all :p 2011-03-20T05:03:49 it's not counted 2011-03-20T05:03:57 he just did 2011-03-20T05:04:04 i used it as own friend in my formulae 2011-03-20T05:04:15 <_flag> sigh: That's better than what we have now because it just leads to people sending in 1 ant at a time 2011-03-20T05:04:21 so one ant always have a friend. 2011-03-20T05:04:41 Stocha: an ant is not its own friend for calculations 2011-03-20T05:04:44 it allows blocking. But let's say that blocking is not a problem. 2011-03-20T05:04:45 <_flag> If that's what you wanted my opinion on 2011-03-20T05:05:09 _flag: i don't think it does 2011-03-20T05:05:38 Only way i see with solving blocking is life or statisticall. 2011-03-20T05:05:45 it gives them a better chance at killing off your one when you have nothing to really lose from going all in, and the scoring makes you want more ants inside the range of the dying ant than enemies 2011-03-20T05:06:00 i originally thought that too :P 2011-03-20T05:06:14 well, all in or no in 2011-03-20T05:06:16 and you don't now ? 2011-03-20T05:06:34 blocking 1 width corridors is fine 2011-03-20T05:06:43 how is it fine ? 2011-03-20T05:06:46 if it's too much of a problem we'll make the map gens not make them 2011-03-20T05:06:50 and how do you prove it's the only case ? 2011-03-20T05:07:04 you didn't find this case in the first place. So you need some proof don't you ? 2011-03-20T05:07:15 <_flag> antimatroid: are you referring to optionn 1 or 2? 2011-03-20T05:07:16 Even i you did find it, you still need a simple proof. 2011-03-20T05:07:18 we did find the 1 corridor case 2011-03-20T05:07:27 _flag: the new option 2011-03-20T05:07:27 and i don't think that such proof will exists :) 2011-03-20T05:07:39 Stocha: challenge accepted 2011-03-20T05:07:46 okay then :) 2011-03-20T05:07:47 but i'm busy atm, sometime in the next week or two 2011-03-20T05:07:52 spit there, and shake hand ! 2011-03-20T05:08:09 <_flag> antimatroid: I was referring to the current one (which I still have as option 2, I don't know if it's been updated) 2011-03-20T05:08:24 _flag: it's about to change 2011-03-20T05:08:40 if you make a mathematically acceptable proof that blocking can only happen in one width corridor, i won't give any more arguments about fighting rules ! 2011-03-20T05:08:50 we'll need a mathematician though 2011-03-20T05:08:54 new method is: your ant dies if it has more enemies in range than friends, not counting itself as a friend 2011-03-20T05:09:07 <_flag> Wait, so what's the difference between the new one and option 1? 2011-03-20T05:09:08 do i count as a mathematician? 2011-03-20T05:09:14 i'm only an honours student 2011-03-20T05:09:21 otherwise i'll pass it on to jmcarthur 2011-03-20T05:09:50 jmcarthur: you're now our resident mathematician :) 2011-03-20T05:09:59 aichallenge: sigh epsilon * r6f19970 / ants/ants.py : Cleanup code a bit - http://bit.ly/gBoywP 2011-03-20T05:10:00 aichallenge: sigh epsilon * r7a961b7 / (ants/ants.py ants/playgame.py): Make support default attack method. Comment do_attack_support. - http://bit.ly/dRU4GE 2011-03-20T05:10:01 you certainly not count as your own proof validator :p 2011-03-20T05:10:19 my friend is a little too keen on actually writing one of those 2011-03-20T05:10:22 you build the proof, and a mathematician check for compleetness 2011-03-20T05:10:25 ok, it's now the default attack method 2011-03-20T05:10:27 from the very bottom up 2011-03-20T05:10:31 well, almost 2011-03-20T05:10:37 we'd extend it to set theory later 2011-03-20T05:10:45 sigh: :) 2011-03-20T05:10:58 now i have to actually work out how to run this on ubuntu 2011-03-20T05:11:17 There is only limited number of case to study. So it might be possible to build up a program for solving the blocking problem also. 2011-03-20T05:11:46 *** kaemo has joined #aichallenge 2011-03-20T05:11:54 when we have come across rules that allowed blocking 2011-03-20T05:12:00 they always happened in a fairly obvious way 2011-03-20T05:12:10 wich challenger no doubt will do once the rule is gold. 2011-03-20T05:12:17 and even then i don't think they would have been easy to do very well 2011-03-20T05:12:27 that i don't argue with. 2011-03-20T05:12:45 only with the fact that there exists some case with blocking. 2011-03-20T05:13:04 antimatroid: McLeopold was in support of it earlier, so I don't think he would have a problem with it being default 2011-03-20T05:13:09 and really, does it matter if people can "make their own walls" 2011-03-20T05:13:21 people just need to not be stupid enough to walk into such a trap 2011-03-20T05:13:44 which is really a nice starting point for writing your own emulate state iteration function 2011-03-20T05:13:50 sigh: neither 2011-03-20T05:13:53 they might have no choice although 2011-03-20T05:13:57 if all ways are blocked 2011-03-20T05:14:02 then they lost i guess 2011-03-20T05:14:09 it's a single game 2011-03-20T05:14:18 but same goes with any rules. Although blocking might be dynamicall 2011-03-20T05:14:19 provided it doesn't end up being the same thing happening every game, i think it's all good 2011-03-20T05:14:41 it's also not trivial to check that you're blocking something 2011-03-20T05:14:50 we might know for quite sure, once the challenge is well under way. 2011-03-20T05:15:10 if blocking is the deciding factor, games may not be fun to watch after a few weeks :p 2011-03-20T05:15:11 (i'm quite good at saying something is trivial when i want it to be and not otherwise, :P) 2011-03-20T05:15:25 when everyone go straight at blocking point, and the randomness does the rest. 2011-03-20T05:15:36 yeah, everyone agrees on that being bad i think 2011-03-20T05:15:52 i agree on it being not fun to watch. 2011-03-20T05:15:59 if that happened we'd almost have to consider changing attack rules, but i really don't think it's possible to block passsages greater than 1 2011-03-20T05:16:04 especially if it happens in the few first days of the challenge 2011-03-20T05:16:06 and we can eliminate those fairly easily if neede 2011-03-20T05:16:09 and then the game is solved :p 2011-03-20T05:16:21 you can't change the rules once the realy thing start. 2011-03-20T05:16:25 that's the problem. 2011-03-20T05:16:28 actually, no it's not 2011-03-20T05:16:32 but it makes for an interesting game 2011-03-20T05:16:38 if you all block off, then no one gets points 2011-03-20T05:16:48 two of you unblocking and fighting will get you above everyone else to win 2011-03-20T05:16:55 aichallenge: sigh epsilon * r452dc87 / ants/ants.py : Stop hard-coding maxx dist in do_attack_closest - http://bit.ly/eNFIWW 2011-03-20T05:16:56 by interesting i mean weird 2011-03-20T05:17:30 well if blocking is paramount, there is bound to be one or two ant dies in the early stages. 2011-03-20T05:17:32 the scoring makes it so that you have to be involved in the game 2011-03-20T05:17:35 not just grow your army 2011-03-20T05:17:37 so that will be the noise. 2011-03-20T05:18:00 you have to be involved in order to grow your army. 2011-03-20T05:18:11 two players could cooperate and just fight in their zoned up blocked region 2011-03-20T05:18:12 and win 2011-03-20T05:18:23